TRIFID

TRIFID classifies splice isoforms to predict their biological significance using machine learning trained on proteomics and genomic features.

Key Features:

• Machine learning classifier: Trained on proteomics-derived evidence to predict which splice isoforms are biologically important.
• Evolutionary features: Uses cross-species conservation metrics to assess isoform preservation and functional constraint.
• Structural features: Evaluates functional domain integrity to detect disruptions caused by alternative splicing.
• Annotation and splicing data: Incorporates genomic annotations and splicing patterns to contextualize isoform production.
• RNA-seq based features: Utilizes transcript abundance from RNA sequencing to inform isoform expression levels.
• Classification criteria: Distinguishes important isoforms by conservation across species, intact functional domains, and evidence of purifying selection on encoding exons.
• Species scope: Developed for the human genome with methodology adaptable to other well-annotated species.

Scientific Applications:

• Isoform prioritization: Identifies splice variants likely to contribute to cellular function based on integrated proteomic, evolutionary, structural, annotation, and expression evidence.
• Variant interpretation: Aids assessment of somatic and germline variants by linking them to predicted functional isoforms.
• Alternative splicing research: Provides evidence to study mechanisms and consequences of alternative splicing in health and disease.

Methodology:

Machine learning model trained on proteomics evidence that integrates cross-species conservation metrics, domain integrity analyses, genomic annotations and splicing patterns, and RNA-seq transcript abundance to evaluate conservation, functional-domain integrity, and purifying selection of exons encoding isoforms.